Sandbox
@taisly/agent

MCP server, CLI, and SDK for video posting

Taisly Agent Kit connects agent workflows to Taisly’s video publishing API. It gives you a CLI, MCP server, SDK, and agent skill so your agent can check accounts, validate a video, confirm with you, publish it, and track status.

219 stars4 forksJavaScriptUpdated 2mo ago
Who it's for

Builders who want their agent to publish short-form videos to connected social accounts.

What it delivers

You can move from a draft video and caption to a published post without wiring every platform integration yourself.

What it does

Agent skill

`SKILL.md` teaches agents the safe posting flow: auth, platforms, validate, confirm, create, status.

CLI commands

`taisly auth:status`, `taisly platforms:list`, `taisly posts:validate`, `taisly posts:create`, and `taisly posts:status` return JSON for automation.

MCP server

Provides local stdio MCP tools and a remote HTTP MCP endpoint for agent tools like setup, platform checks, validation, posting, and status.

Codex and Claude plugin manifests

`.codex-plugin/plugin.json` and `.claude-plugin/plugin.json` point agents to the skill and MCP server.

SDK for scripts

The JavaScript SDK exposes `platforms.list()`, `posts.validate()`, and `posts.create()` for custom workflows.

Copy-paste examples

`examples/` includes workflows for Codex, Claude Code, Cursor, and shell scripts for posting and scheduling.

How to get it

  1. 1Run
    npm install -g @taisly/agent
  2. 2Or run it without a global install
    npx @taisly/agent help
  3. 3Start the stdio MCP server
    npx @taisly/agent mcp
  4. 4For local development inside this repository
    node packages/agent/src/cli.js help
  5. 5List the connected social accounts available to the API key
    taisly auth:status
    taisly platforms:list
  6. 6Connect a social account from an AI-agent flow
    taisly platforms:connect:start --platform instagram
    # Open the returned connectUrl and finish authorization in the browser.
    taisly platforms:connect:check --platform instagram
    taisly platforms:list

README

Taisly Agent Kit: AI agent social media posting for TikTok, Instagram Reels, YouTube Shorts, X, and Facebook

Taisly Agent Kit

Taisly MCP server Taisly MCP score

Free AI-first social media posting for short-form video. Taisly Agent Kit is a JSON-first SDK, CLI, Agent Skill, and MCP server that lets AI agents, developer tools, and automation workflows publish videos to TikTok, Instagram Reels, YouTube Shorts, X, Facebook, and other connected social platforms through Taisly.

Use it when your agent can create content, write captions, or prepare a campaign, but still needs a reliable video publishing API to put that content online. The package is free to install and designed to get an agent from idea to published video without building every platform integration yourself.

Website | Agent Kit | API Docs | npm | GitHub | MCP Registry | Glama | Guide

Languages

Read this guide in: English, Español, Deutsch, Français, Português, ไทย, 中文, Bahasa Indonesia, Русский, Nederlands, 한국어, 日本語, العربية, Türkçe, Polski.

Why developers use it

  • Build AI agent social media posting into Codex, Claude Code, Cursor, OpenClaw, and custom automation tools.
  • Automate video publishing from one local file to multiple connected social accounts.
  • Give agents a safe JSON workflow: discover accounts, validate payloads, create posts, and check status.
  • Add TikTok API posting automation, Instagram Reels automation, YouTube Shorts publishing, and cross-platform social media automation without building every platform integration yourself.
  • Start free, then upgrade only when the publishing workflow grows.

Taisly handles the connected accounts and posting execution. Your agent handles planning, caption writing, campaign logic, or workflow orchestration.

Free AI-first posting

  • Install the package for free.
  • Create or connect a Taisly account from your AI agent.
  • Let the agent authenticate, connect social accounts, validate videos, and publish through a JSON-first workflow.
  • Upgrade later when you need higher-volume publishing, larger account sets, or automatic repost workflows.

No credit card is required to get started. Paid plans are for teams and workflows that need more scale.

Agent Skill

This package includes the Taisly Social Media Posting Skill in SKILL.md. Use it with Claude Code, Codex, Cursor, OpenClaw, and other agents when you want the agent to understand the safe posting workflow before it touches live social accounts.

Recommended skill workflow:

auth -> platforms -> validate -> confirm -> create -> status

The skill tells agents to discover connected accounts, validate the post, ask for explicit user confirmation, create the post, and save the returned historyId for status checks.

Codex plugin

This package includes a Codex plugin manifest in .codex-plugin/plugin.json. The plugin points Codex to the canonical SKILL.md file and the official remote MCP server:

https://app.taisly.com/mcp

Use the plugin when you want Codex to discover both the posting workflow instructions and the Taisly MCP tools from one package. The root SKILL.md is the single source of truth for the agent workflow.

Install

npm install -g @taisly/agent

Or run it without a global install:

npx @taisly/agent help

Start the stdio MCP server:

npx @taisly/agent mcp

For local development inside this repository:

node packages/agent/src/cli.js help

Authentication

The easiest flow is to let your AI agent open the browser login for you. Use any short agent slug, for example claude-code, codex, cursor, windsurf, openclaw, hermes-agent, cline, or aider.

Help me set up Taisly Agent Kit.

1. Install the CLI: npm i -g @taisly/agent
2. Run taisly setup --agent <agent-slug> and send me the login URL so I can finish authentication.
3. Wait for me to finish login in the browser.
4. Run taisly checkin --agent <agent-slug>.
5. Tell me if Taisly is connected and ready. If not, tell me what failed.

If you cannot install or run the Taisly CLI in this environment, stop and tell me to open Taisly Settings instead.

The setup command stores a local Taisly agent credential, so later CLI and MCP commands can run without manually copying an API key.

Manual fallback: create an API key in Taisly Settings, then set:

export TAISLY_API_KEY="taisly_..."

Advanced: by default, the browser setup stores the local credential in ~/.taisly/config.json. Most users should not change this. Set TAISLY_CONFIG_HOME=/path/to/dir only for CI, tests, or isolated agent profiles that need a separate credential store.

Quick start

List the connected social accounts available to the API key:

taisly auth:status
taisly platforms:list

Connect a social account from an AI-agent flow:

taisly platforms:connect:start --platform instagram
# Open the returned connectUrl and finish authorization in the browser.
taisly platforms:connect:check --platform instagram
taisly platforms:list

Supported connect slugs: instagram, tiktok, youtube, x, facebook.

Validate a local video before publishing:

taisly posts:validate \
  --video ./launch.mp4 \
  --platforms platform_id_1,platform_id_2 \
  --description "Launch day"

Publish now:

taisly posts:create \
  --video ./launch.mp4 \
  --platforms platform_id_1,platform_id_2 \
  --description "Launch day"

Schedule for later:

taisly posts:create \
  --video ./launch.mp4 \
  --platforms platform_id_1,platform_id_2 \
  --description "Launch day" \
  --scheduled "2026-06-14T09:00:00+07:00"

Check the returned post:

taisly posts:status --id <historyId>

Every command prints JSON so agents can parse results without scraping terminal text.

MCP server

Taisly is published in the official MCP Registry as:

io.github.taisly/agent

Display name: Taisly Social Media Posting

Remote MCP endpoint:

https://app.taisly.com/mcp

The remote MCP server uses streamable HTTP and OAuth. Remote MCP tools can publish only from public videoUrl values because they cannot read local files from your machine.

Codex remote MCP setup:

codex mcp add taisly --url https://app.taisly.com/mcp
codex mcp login taisly

Claude Code remote MCP setup:

claude mcp add --transport http taisly https://app.taisly.com/mcp

Then ask Claude Code to verify the connection. If OAuth is needed, Claude Code can guide the browser login from the session. You can also authenticate explicitly with claude mcp login taisly if you want to complete the OAuth flow from the shell before using tools.

Taisly Agent Kit also includes a local stdio MCP server in the same package. Use local MCP when your agent needs to read local video files:

{
  "mcpServers": {
    "taisly": {
      "command": "npx",
      "args": ["@taisly/agent", "mcp"]
    }
  }
}

If you already use a manual API key, you can still pass TAISLY_API_KEY in the MCP server env block.

Codex local MCP setup:

codex mcp add taisly -- npx @taisly/agent mcp

Claude Code local MCP setup:

claude mcp add taisly -- npx @taisly/agent mcp

Then ask the agent:

Use the local Taisly MCP server you just added and run the Taisly agent setup start tool with agentId: "<agent-slug>".

Send me the login URL and wait for me to finish in the browser.
Then run the Taisly agent checkin tool with agentId: "<agent-slug>".
After checkin, run the Taisly auth status tool.
Tell me if Taisly is connected and ready. If not, tell me what failed.

Available MCP tools:

  • taisly_agent_setup_start
  • taisly_agent_checkin
  • taisly_auth_status
  • taisly_platforms_list
  • taisly_platform_schema
  • taisly_platform_connect_start
  • taisly_platform_connect_check
  • taisly_posts_validate
  • taisly_posts_create
  • taisly_posts_status
  • taisly_posts_list
  • taisly_reposts_list
  • taisly_reposts_create

taisly_posts_create requires confirmed: true. Set it only after the user explicitly approves the video, destination accounts, caption, and schedule.

JSON workflow for agents

Agents can write a payload file and pass it to the CLI:

{
  "video": "./launch.mp4",
  "platforms": ["platform_id_1", "platform_id_2"],
  "description": "Launch day. Short demo, big update.",
  "scheduled": "2026-06-14T09:00:00+07:00"
}

Then run:

taisly posts:validate --json ./campaign.json
taisly posts:create --json ./campaign.json

The video path must point to a real local file available to the agent. Supported local preflight extensions are .mp4, .mov, .avi, .mkv, .webm, .flv, .mpeg, and .mpg.

Commands

taisly auth:status
taisly setup --agent <agent-slug>
taisly checkin --agent <agent-slug>
taisly platforms:list
taisly integrations:list
taisly platforms:schema --platform TikTok
taisly posts:validate --video ./launch.mp4 --platforms platform_id_1 --description "Launch day"
taisly posts:create --video ./launch.mp4 --platforms platform_id_1 --description "Launch day"
taisly posts:create --json ./campaign.json
taisly posts:list --page 1
taisly posts:status --id <historyId>
taisly reposts:list
taisly reposts:create --from <platform_id> --to <platform_id_1,platform_id_2>
taisly mcp

integrations:* commands are aliases for platforms:* commands. Taisly calls connected social accounts platforms in the app, while many public APIs call them integrations.

SDK

import { Taisly } from "@taisly/agent";

const taisly = new Taisly({
  apiKey: process.env.TAISLY_API_KEY,
});

const platforms = await taisly.platforms.list();

const validation = await taisly.posts.validate({
  video: "./launch.mp4",
  platforms: [platforms.data[0].id],
  description: "Launch day",
});

const post = await taisly.posts.create({
  video: "./launch.mp4",
  platforms: [platforms.data[0].id],
  description: "Launch day",
  scheduled: "2026-06-14T09:00:00+07:00",
});

console.log(validation.success);
console.log(post.historyId);

Use with AI agents

Taisly Agent Kit is designed for agentic workflows where the user gives a high-level instruction and the agent executes a safe posting path.

  • Codex: let a coding agent publish demo videos, launch clips, or build updates after user confirmation.
  • Claude Code: add social posting to local content workflows and campaign scripts.
  • Cursor: ship video publishing from developer tools, content apps, and internal automation.
  • OpenClaw and other agents: connect planning, caption writing, and posting execution through a single CLI.

Recommended agent path:

auth:status -> platforms:list -> platforms:schema -> posts:validate -> user confirmation -> posts:create -> posts:status

Agent recipes

The examples/ folder includes copy-paste workflows for common coding agents:

  • examples/codex/post-video.md
  • examples/claude-code/schedule-video.md
  • examples/cursor/post-build-demo-video.md
  • examples/post-video.sh
  • examples/schedule-video.sh

What this package is not

  • It is not a social media dashboard.
  • It is not a video editor.
  • It does not replace a Taisly account.
  • It requires connected social accounts and a Taisly agent credential or API key.
  • It does not bypass platform rules, account permissions, or media validation.

Current limits

  • posts:create uses the existing multipart /post API.
  • posts:status reads recent history because a single-post status endpoint is not available yet.
  • posts:validate is local preflight; final validation still happens in Taisly.
  • Local MCP uses stdio transport. Remote MCP uses streamable HTTP at https://app.taisly.com/mcp.
  • Remote MCP requires public video URLs; use local MCP or the CLI for local video files.
  • Media upload reuse is planned later.

Links

Files in the repo

Repository payload15 top-level entries
  • .claude-plugin
  • .codex-plugin
  • .github
  • assets
  • docs
  • examples
  • src
  • .gitignore
  • Dockerfile
  • glama.json
  • LICENSE
  • package.json
  • README.md
  • server.json
  • SKILL.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More connectors

Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface

86k

High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.

43k

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code

14k
okf-memory/
okf-agent-memory

Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.

547
tirth8205/
code-review-graph

Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.

31k
2akouwu/
reverify

Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.

1.1k